commit a3e2ba88eb09c1eed2f7ed6e17660b345464bb90 Author: law Date: Wed Sep 20 05:05:12 2017 +0000 2017-09-18 Jeff Law * explow.c: Include "params.h" and "dumpfile.h". (anti_adjust_stack_and_probe_stack_clash): New function. (get_stack_check_protect): Likewise. (compute_stack_clash_protection_loop_data): Likewise. (emit_stack_clash_protection_loop_start): Likewise. (emit_stack_clash_protection_loop_end): Likewise. (allocate_dynamic_stack_space): Use get_stack_check_protect. Use anti_adjust_stack_and_probe_stack_clash. * explow.h (compute_stack_clash_protection_loop_data): Prototype. (emit_stack_clash_protection_loop_start): Likewise. (emit_stack_clash_protection_loop_end): Likewise. * rtl.h (get_stack_check_protect): Prototype. * target.def (stack_clash_protection_final_dynamic_probe): New hook. * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New. * targhooks.h (default_stack_clash_protection_final_dynamic_probe): Prototype. * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Add @hook. * doc/tm.texi: Rebuilt. * config/alpha/alpha.c (alpha_expand_prologue): Likewise. * config/i386/i386.c (ix86_expand_prologue): Likewise. * config/ia64/ia64.c (ia64_expand_prologue): Likewise. * config/mips/mips.c (mips_expand_prologue): Likewise. * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise. * config/sparc/sparc.c (sparc_expand_prologue): Likewise. (sparc_flat_expand_prologue): Likewise. * gcc.dg/stack-check-3.c: New test. git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@252995 138bc75d-0d04-0410-961f-82ee72b054a4 diff --git a/gcc/config/alpha/alpha.c b/gcc/config/alpha/alpha.c index 2874b8454a9..5402f5213d6 100644 --- a/gcc/config/alpha/alpha.c +++ b/gcc/config/alpha/alpha.c @@ -7625,7 +7625,7 @@ alpha_expand_prologue (void) probed_size = frame_size; if (flag_stack_check) - probed_size += STACK_CHECK_PROTECT; + probed_size += get_stack_check_protect (); if (probed_size <= 32768) { diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c index e36726ba722..d996fd160e8 100644 --- a/gcc/config/i386/i386.c +++ b/gcc/config/i386/i386.c @@ -10544,12 +10544,12 @@ ix86_expand_prologue (void) HOST_WIDE_INT size = allocate; if (TARGET_64BIT && size >= (HOST_WIDE_INT) 0x80000000) - size = 0x80000000 - STACK_CHECK_PROTECT - 1; + size = 0x80000000 - get_stack_check_protect () - 1; if (TARGET_STACK_PROBE) - ix86_emit_probe_stack_range (0, size + STACK_CHECK_PROTECT); + ix86_emit_probe_stack_range (0, size + get_stack_check_protect ()); else - ix86_emit_probe_stack_range (STACK_CHECK_PROTECT, size); + ix86_emit_probe_stack_range (get_stack_check_protect (), size); } } diff --git a/gcc/config/ia64/ia64.c b/gcc/config/ia64/ia64.c index 50bbad6661c..390983936e8 100644 --- a/gcc/config/ia64/ia64.c +++ b/gcc/config/ia64/ia64.c @@ -3435,7 +3435,7 @@ ia64_expand_prologue (void) current_function_static_stack_size = current_frame_info.total_size; if (flag_stack_check == STATIC_BUILTIN_STACK_CHECK) - ia64_emit_probe_stack_range (STACK_CHECK_PROTECT, + ia64_emit_probe_stack_range (get_stack_check_protect (), current_frame_info.total_size, current_frame_info.n_input_regs + current_frame_info.n_local_regs); diff --git a/gcc/config/mips/mips.c b/gcc/config/mips/mips.c index 41c5d6b6b1f..9b7eb678f19 100644 --- a/gcc/config/mips/mips.c +++ b/gcc/config/mips/mips.c @@ -10746,7 +10746,7 @@ mips_expand_prologue (void) current_function_static_stack_size = size; if (flag_stack_check == STATIC_BUILTIN_STACK_CHECK && size) - mips_emit_probe_stack_range (STACK_CHECK_PROTECT, size); + mips_emit_probe_stack_range (get_stack_check_protect (), size); /* Save the registers. Allocate up to MIPS_MAX_FIRST_STACK_STEP bytes beforehand; this is enough to cover the register save area diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c index 15583055895..a9052c6becf 100644 --- a/gcc/config/rs6000/rs6000.c +++ b/gcc/config/rs6000/rs6000.c @@ -23214,7 +23214,8 @@ rs6000_emit_prologue (void) current_function_static_stack_size = info->total_size; if (flag_stack_check == STATIC_BUILTIN_STACK_CHECK && info->total_size) - rs6000_emit_probe_stack_range (STACK_CHECK_PROTECT, info->total_size); + rs6000_emit_probe_stack_range (get_stack_check_protect (), + info->total_size); if (TARGET_FIX_AND_CONTINUE) { diff --git a/gcc/config/sparc/sparc.c b/gcc/config/sparc/sparc.c index e5d326cdf23..e5e93c80261 100644 --- a/gcc/config/sparc/sparc.c +++ b/gcc/config/sparc/sparc.c @@ -5431,7 +5431,7 @@ sparc_expand_prologue (void) current_function_static_stack_size = size; if (flag_stack_check == STATIC_BUILTIN_STACK_CHECK && size) - sparc_emit_probe_stack_range (STACK_CHECK_PROTECT, size); + sparc_emit_probe_stack_range (get_stack_check_protect (), size); if (size == 0) ; /* do nothing. */ @@ -5533,7 +5533,7 @@ sparc_flat_expand_prologue (void) current_function_static_stack_size = size; if (flag_stack_check == STATIC_BUILTIN_STACK_CHECK && size) - sparc_emit_probe_stack_range (STACK_CHECK_PROTECT, size); + sparc_emit_probe_stack_range (get_stack_check_protect (), size); if (sparc_save_local_in_regs_p) emit_save_or_restore_local_in_regs (stack_pointer_rtx, SPARC_STACK_BIAS, diff --git a/gcc/doc/tm.texi b/gcc/doc/tm.texi index 6b18a2724bc..eeef757bf5b 100644 --- a/gcc/doc/tm.texi +++ b/gcc/doc/tm.texi @@ -3571,6 +3571,10 @@ GCC computed the default from the values of the above macros and you will normally not need to override that default. @end defmac +@deftypefn {Target Hook} bool TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE (rtx @var{residual}) +Some targets make optimistic assumptions about the state of stack probing when they emit their prologues. On such targets a probe into the end of any dynamically allocated space is likely required for safety against stack clash style attacks. Define this variable to return nonzero if such a probe is required or zero otherwise. You need not define this macro if it would always have the value zero. +@end deftypefn + @need 2000 @node Frame Registers @subsection Registers That Address the Stack Frame diff --git a/gcc/doc/tm.texi.in b/gcc/doc/tm.texi.in index 7d0b3c73b2f..6707ca87236 100644 --- a/gcc/doc/tm.texi.in +++ b/gcc/doc/tm.texi.in @@ -3539,6 +3539,8 @@ GCC computed the default from the values of the above macros and you will normally not need to override that default. @end defmac +@hook TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE + @need 2000 @node Frame Registers @subsection Registers That Address the Stack Frame diff --git a/gcc/explow.c b/gcc/explow.c index 7da8bc75f19..2526e8513b7 100644 --- a/gcc/explow.c +++ b/gcc/explow.c @@ -40,8 +40,11 @@ along with GCC; see the file COPYING3. If not see #include "target.h" #include "common/common-target.h" #include "output.h" +#include "params.h" +#include "dumpfile.h" static rtx break_out_memory_refs (rtx); +static void anti_adjust_stack_and_probe_stack_clash (rtx); /* Truncate and perhaps sign-extend C as appropriate for MODE. */ @@ -1140,6 +1143,29 @@ update_nonlocal_goto_save_area (void) emit_stack_save (SAVE_NONLOCAL, &r_save); } +/* Return the number of bytes to "protect" on the stack for -fstack-check. + + "protect" in the context of -fstack-check means how many bytes we + should always ensure are available on the stack. More importantly + this is how many bytes are skipped when probing the stack. + + On some targets we want to reuse the -fstack-check prologue support + to give a degree of protection against stack clashing style attacks. + + In that scenario we do not want to skip bytes before probing as that + would render the stack clash protections useless. + + So we never use STACK_CHECK_PROTECT directly. Instead we indirect though + this helper which allows us to provide different values for + -fstack-check and -fstack-clash-protection. */ +HOST_WIDE_INT +get_stack_check_protect (void) +{ + if (flag_stack_clash_protection) + return 0; + return STACK_CHECK_PROTECT; +} + /* Return an rtx representing the address of an area of memory dynamically pushed on the stack. @@ -1393,7 +1419,7 @@ allocate_dynamic_stack_space (rtx size, unsigned size_align, probe_stack_range (STACK_OLD_CHECK_PROTECT + STACK_CHECK_MAX_FRAME_SIZE, size); else if (flag_stack_check == STATIC_BUILTIN_STACK_CHECK) - probe_stack_range (STACK_CHECK_PROTECT, size); + probe_stack_range (get_stack_check_protect (), size); /* Don't let anti_adjust_stack emit notes. */ suppress_reg_args_size = true; @@ -1451,6 +1477,8 @@ allocate_dynamic_stack_space (rtx size, unsigned size_align, if (flag_stack_check && STACK_CHECK_MOVING_SP) anti_adjust_stack_and_probe (size, false); + else if (flag_stack_clash_protection) + anti_adjust_stack_and_probe_stack_clash (size); else anti_adjust_stack (size); @@ -1712,6 +1740,219 @@ probe_stack_range (HOST_WIDE_INT first, rtx size) } } +/* Compute parameters for stack clash probing a dynamic stack + allocation of SIZE bytes. + + We compute ROUNDED_SIZE, LAST_ADDR, RESIDUAL and PROBE_INTERVAL. + + Additionally we conditionally dump the type of probing that will + be needed given the values computed. */ + +void +compute_stack_clash_protection_loop_data (rtx *rounded_size, rtx *last_addr, + rtx *residual, + HOST_WIDE_INT *probe_interval, + rtx size) +{ + /* Round SIZE down to STACK_CLASH_PROTECTION_PROBE_INTERVAL */ + *probe_interval + = 1 << PARAM_VALUE (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL); + *rounded_size = simplify_gen_binary (AND, Pmode, size, + GEN_INT (-*probe_interval)); + + /* Compute the value of the stack pointer for the last iteration. + It's just SP + ROUNDED_SIZE. */ + rtx rounded_size_op = force_operand (*rounded_size, NULL_RTX); + *last_addr = force_operand (gen_rtx_fmt_ee (STACK_GROW_OP, Pmode, + stack_pointer_rtx, + rounded_size_op), + NULL_RTX); + + /* Compute any residuals not allocated by the loop above. Residuals + are just the ROUNDED_SIZE - SIZE. */ + *residual = simplify_gen_binary (MINUS, Pmode, size, *rounded_size); + + /* Dump key information to make writing tests easy. */ + if (dump_file) + { + if (*rounded_size == CONST0_RTX (Pmode)) + fprintf (dump_file, + "Stack clash skipped dynamic allocation and probing loop.\n"); + else if (GET_CODE (*rounded_size) == CONST_INT + && INTVAL (*rounded_size) <= 4 * *probe_interval) + fprintf (dump_file, + "Stack clash dynamic allocation and probing inline.\n"); + else if (GET_CODE (*rounded_size) == CONST_INT) + fprintf (dump_file, + "Stack clash dynamic allocation and probing in " + "rotated loop.\n"); + else + fprintf (dump_file, + "Stack clash dynamic allocation and probing in loop.\n"); + + if (*residual != CONST0_RTX (Pmode)) + fprintf (dump_file, + "Stack clash dynamic allocation and probing residuals.\n"); + else + fprintf (dump_file, + "Stack clash skipped dynamic allocation and " + "probing residuals.\n"); + } +} + +/* Emit the start of an allocate/probe loop for stack + clash protection. + + LOOP_LAB and END_LAB are returned for use when we emit the + end of the loop. + + LAST addr is the value for SP which stops the loop. */ +void +emit_stack_clash_protection_probe_loop_start (rtx *loop_lab, + rtx *end_lab, + rtx last_addr, + bool rotated) +{ + /* Essentially we want to emit any setup code, the top of loop + label and the comparison at the top of the loop. */ + *loop_lab = gen_label_rtx (); + *end_lab = gen_label_rtx (); + + emit_label (*loop_lab); + if (!rotated) + emit_cmp_and_jump_insns (stack_pointer_rtx, last_addr, EQ, NULL_RTX, + Pmode, 1, *end_lab); +} + +/* Emit the end of a stack clash probing loop. + + This consists of just the jump back to LOOP_LAB and + emitting END_LOOP after the loop. */ + +void +emit_stack_clash_protection_probe_loop_end (rtx loop_lab, rtx end_loop, + rtx last_addr, bool rotated) +{ + if (rotated) + emit_cmp_and_jump_insns (stack_pointer_rtx, last_addr, NE, NULL_RTX, + Pmode, 1, loop_lab); + else + emit_jump (loop_lab); + + emit_label (end_loop); + +} + +/* Adjust the stack pointer by minus SIZE (an rtx for a number of bytes) + while probing it. This pushes when SIZE is positive. SIZE need not + be constant. + + This is subtly different than anti_adjust_stack_and_probe to try and + prevent stack-clash attacks + + 1. It must assume no knowledge of the probing state, any allocation + must probe. + + Consider the case of a 1 byte alloca in a loop. If the sum of the + allocations is large, then this could be used to jump the guard if + probes were not emitted. + + 2. It never skips probes, whereas anti_adjust_stack_and_probe will + skip probes on the first couple PROBE_INTERVALs on the assumption + they're done elsewhere. + + 3. It only allocates and probes SIZE bytes, it does not need to + allocate/probe beyond that because this probing style does not + guarantee signal handling capability if the guard is hit. */ + +static void +anti_adjust_stack_and_probe_stack_clash (rtx size) +{ + /* First ensure SIZE is Pmode. */ + if (GET_MODE (size) != VOIDmode && GET_MODE (size) != Pmode) + size = convert_to_mode (Pmode, size, 1); + + /* We can get here with a constant size on some targets. */ + rtx rounded_size, last_addr, residual; + HOST_WIDE_INT probe_interval; + compute_stack_clash_protection_loop_data (&rounded_size, &last_addr, + &residual, &probe_interval, size); + + if (rounded_size != CONST0_RTX (Pmode)) + { + if (INTVAL (rounded_size) <= 4 * probe_interval) + { + for (HOST_WIDE_INT i = 0; + i < INTVAL (rounded_size); + i += probe_interval) + { + anti_adjust_stack (GEN_INT (probe_interval)); + + /* The prologue does not probe residuals. Thus the offset + here to probe just beyond what the prologue had already + allocated. */ + emit_stack_probe (plus_constant (Pmode, stack_pointer_rtx, + (probe_interval + - GET_MODE_SIZE (word_mode)))); + emit_insn (gen_blockage ()); + } + } + else + { + rtx loop_lab, end_loop; + bool rotate_loop = GET_CODE (rounded_size) == CONST_INT; + emit_stack_clash_protection_probe_loop_start (&loop_lab, &end_loop, + last_addr, rotate_loop); + + anti_adjust_stack (GEN_INT (probe_interval)); + + /* The prologue does not probe residuals. Thus the offset here + to probe just beyond what the prologue had already allocated. */ + emit_stack_probe (plus_constant (Pmode, stack_pointer_rtx, + (probe_interval + - GET_MODE_SIZE (word_mode)))); + + emit_stack_clash_protection_probe_loop_end (loop_lab, end_loop, + last_addr, rotate_loop); + emit_insn (gen_blockage ()); + } + } + + if (residual != CONST0_RTX (Pmode)) + { + rtx x = force_reg (Pmode, plus_constant (Pmode, residual, + -GET_MODE_SIZE (word_mode))); + anti_adjust_stack (residual); + emit_stack_probe (gen_rtx_PLUS (Pmode, stack_pointer_rtx, x)); + emit_insn (gen_blockage ()); + } + + /* Some targets make optimistic assumptions in their prologues about + how the caller may have probed the stack. Make sure we honor + those assumptions when needed. */ + if (size != CONST0_RTX (Pmode) + && targetm.stack_clash_protection_final_dynamic_probe (residual)) + { + /* Ideally we would just probe at *sp. However, if SIZE is not + a compile-time constant, but is zero at runtime, then *sp + might hold live data. So probe at *sp if we know that + an allocation was made, otherwise probe into the red zone + which is obviously undesirable. */ + if (GET_CODE (size) == CONST_INT) + { + emit_stack_probe (stack_pointer_rtx); + emit_insn (gen_blockage ()); + } + else + { + emit_stack_probe (plus_constant (Pmode, stack_pointer_rtx, + -GET_MODE_SIZE (word_mode))); + emit_insn (gen_blockage ()); + } + } +} + + /* Adjust the stack pointer by minus SIZE (an rtx for a number of bytes) while probing it. This pushes when SIZE is positive. SIZE need not be constant. If ADJUST_BACK is true, adjust back the stack pointer diff --git a/gcc/rtl.h b/gcc/rtl.h index 91f3387c701..ab8ec27418d 100644 --- a/gcc/rtl.h +++ b/gcc/rtl.h @@ -1756,6 +1756,17 @@ extern int currently_expanding_to_rtl; /* In explow.c */ extern HOST_WIDE_INT trunc_int_for_mode (HOST_WIDE_INT, enum machine_mode); extern rtx plus_constant (enum machine_mode, rtx, HOST_WIDE_INT); +extern HOST_WIDE_INT get_stack_check_protect (void); + +/* Support for building allocation/probing loops for stack-clash + protection of dyamically allocated stack space. */ +extern void compute_stack_clash_protection_loop_data (rtx *, rtx *, rtx *, + HOST_WIDE_INT *, rtx); +extern void emit_stack_clash_protection_probe_loop_start (rtx *, rtx *, + rtx, bool); +extern void emit_stack_clash_protection_probe_loop_end (rtx, rtx, + rtx, bool); + /* In rtl.c */ extern rtx rtx_alloc_stat (RTX_CODE MEM_STAT_DECL); diff --git a/gcc/target.def b/gcc/target.def index 4d6081c3121..eb2bd46f7a1 100644 --- a/gcc/target.def +++ b/gcc/target.def @@ -2580,6 +2580,13 @@ DEFHOOK void, (void), hook_void_void) +DEFHOOK +(stack_clash_protection_final_dynamic_probe, + "Some targets make optimistic assumptions about the state of stack probing when they emit their prologues. On such targets a probe into the end of any dynamically allocated space is likely required for safety against stack clash style attacks. Define this variable to return nonzero if such a probe is required or zero otherwise. You need not define this macro if it would always have the value zero.", + bool, (rtx residual), + default_stack_clash_protection_final_dynamic_probe) + + /* Functions specific to the C family of frontends. */ #undef HOOK_PREFIX #define HOOK_PREFIX "TARGET_C_" diff --git a/gcc/targhooks.c b/gcc/targhooks.c index f6aa9907225..be23875538d 100644 --- a/gcc/targhooks.c +++ b/gcc/targhooks.c @@ -1557,4 +1557,10 @@ default_canonicalize_comparison (int *, rtx *, rtx *, bool) { } +bool +default_stack_clash_protection_final_dynamic_probe (rtx residual ATTRIBUTE_UNUSED) +{ + return 0; +} + #include "gt-targhooks.h" diff --git a/gcc/targhooks.h b/gcc/targhooks.h index b64274d3ff9..4acf33fae08 100644 --- a/gcc/targhooks.h +++ b/gcc/targhooks.h @@ -195,3 +195,4 @@ extern const char *default_pch_valid_p (const void *, size_t); extern void default_asm_output_ident_directive (const char*); extern bool default_member_type_forces_blk (const_tree, enum machine_mode); +extern bool default_stack_clash_protection_final_dynamic_probe (rtx); diff --git a/gcc/testsuite/gcc.dg/stack-check-3.c b/gcc/testsuite/gcc.dg/stack-check-3.c new file mode 100644 index 00000000000..58fb65649ee --- /dev/null +++ b/gcc/testsuite/gcc.dg/stack-check-3.c @@ -0,0 +1,86 @@ +/* The goal here is to ensure that dynamic allocations via vlas or + alloca calls receive probing. + + Scanning the RTL or assembly code seems like insanity here as does + checking for particular allocation sizes and probe offsets. For + now we just verify that there's an allocation + probe loop and + residual allocation + probe for f?. */ + +/* { dg-do compile } */ +/* { dg-options "-O2 -fstack-clash-protection -fdump-rtl-expand -fno-optimize-sibling-calls --param stack-clash-protection-probe-interval=4096 --param stack-clash-protection-guard-size=4096" } */ +/* { dg-require-effective-target supports_stack_clash_protection } */ + +__attribute__((noinline, noclone)) void +foo (char *p) +{ + asm volatile ("" : : "r" (p) : "memory"); +} + +/* Simple VLA, no other locals. */ +__attribute__((noinline, noclone)) void +f0 (int x) +{ + char vla[x]; + foo (vla); +} + +/* Simple VLA, small local frame. */ +__attribute__((noinline, noclone)) void +f1 (int x) +{ + char locals[128]; + char vla[x]; + foo (vla); +} + +/* Small constant alloca, no other locals. */ +__attribute__((noinline, noclone)) void +f2 (int x) +{ + char *vla = __builtin_alloca (128); + foo (vla); +} + +/* Big constant alloca, small local frame. */ +__attribute__((noinline, noclone)) void +f3 (int x) +{ + char locals[128]; + char *vla = __builtin_alloca (16384); + foo (vla); +} + +/* Big constant alloca, small local frame. */ +__attribute__((noinline, noclone)) void +f3a (int x) +{ + char locals[128]; + char *vla = __builtin_alloca (32768); + foo (vla); +} + +/* Nonconstant alloca, no other locals. */ +__attribute__((noinline, noclone)) void +f4 (int x) +{ + char *vla = __builtin_alloca (x); + foo (vla); +} + +/* Nonconstant alloca, small local frame. */ +__attribute__((noinline, noclone)) void +f5 (int x) +{ + char locals[128]; + char *vla = __builtin_alloca (x); + foo (vla); +} + +/* { dg-final { scan-rtl-dump-times "allocation and probing residuals" 7 "expand" } } */ + + +/* { dg-final { scan-rtl-dump-times "allocation and probing in loop" 7 "expand" { target callee_realigns_stack } } } */ +/* { dg-final { scan-rtl-dump-times "allocation and probing in loop" 4 "expand" { target { ! callee_realigns_stack } } } } */ +/* { dg-final { scan-rtl-dump-times "allocation and probing in rotated loop" 1 "expand" { target { ! callee_realigns_stack } } } } */ +/* { dg-final { scan-rtl-dump-times "allocation and probing inline" 1 "expand" { target { ! callee_realigns_stack } } } } */ +/* { dg-final { scan-rtl-dump-times "skipped dynamic allocation and probing loop" 1 "expand" { target { ! callee_realigns_stack } } } } */